Description of problem: The forcedeth.c version 0.42-rh which is included in 2.6.9-55.0.2 does not support the "sideband management unit" feature which is used on some modern platforms to provide network connectivity for the IPMI BMC. The symptom is that once the admin issues "ifconfig eth0 down" and/or "rmmod forcedeth", the IPMI BMC loses access to the network. The forcedeth.c "0.62-Driver Package V1.23" available at http://www.nvidia.com/object/linux_nforce_1.23.html does support the sideband management unit, as well as providing support for a bunch of new hardware. Version-Release number of selected component (if applicable): kernel 2.6.9-55.0.2 forcedeth.c "0.42-rh" How reproducible: always Steps to Reproduce: 1. install RHEL4u5 on Opteron with MCP55 chipset and IPMI-on-eth0 hardware config, such as Penguin Computing Altus 600. 2. configure IPMI for network accessibility. 3. issue "ifconfig eth0 down; rmmod forcedeth" as root. Actual results: attempts to remotely reboot server using "ipmitool -H 10.0.0.2 chassis power reset" will fail, stating inability to contact BMC. Expected results: remote reset with "ipmitool -H" should work no matter the state of the target OS. Additional info:
The next RHEL4 update will include a later version of the forcedeth driver (0.60). Test kernels that contain a patch for this (and a few other test patches) are available here: http://people.redhat.com/agospoda/#rhel4 Feel free to test these kernels and let us know if they resolve your issue. Thanks!
Adding Ayaz Abdulla to cc list.
restoring needinfo, hadn't meant to remove that. Will try to test soon.
I've verified that 2.6.9-55.21.EL.gtest.26.1 with forcedeth 0.60 fixes the IPMI sideband issue, and it also addresses a locking bug that had been problematic for us with nvidia's 0.60 and 0.62. It appears that you'll want to merge 3ba4d093fe8a26f5f2da94411bf8732fa6e9da86 from upstream as well. Note that 0.62 "Driver Package 1.23" is now available from nvidia.com. Thanks for the great work! This driver looks much better now.
Glad to hear that is working for you, Andy! The last commit I pulled into this update was actually: commit 6fedae1f6e66ab5f169bf58064e23e015fc1307d Author: Ayaz Abdulla <aabdulla> Date: Tue Feb 20 03:34:44 2007 -0500 So I pulled before the one you recommended. commit 3ba4d093fe8a26f5f2da94411bf8732fa6e9da86 Author: Ayaz Abdulla <aabdulla> Date: Fri Mar 23 05:50:02 2007 -0500 It probably won't make the upcoming update (4.6), but I'll pull together some of the latest upstream fixes for my test kernels and hopefully push them into 4.7. If it's OK with you I'd like to close out this bug since your main IPMI issue seems to be resolved. Seem reasonable?
(In reply to comment #5) > If it's OK with you I'd like to close out this bug since your main IPMI issue > seems to be resolved. Seem reasonable? Sounds good to me, but I'd like to make sure 3ba4d09 doesn't fall through the cracks (I'll be maintaining it locally until it shows up in RHEL4.7 or whatever). Would you like to open a new ticket or shall I?
Closing since the IPMI issue is resolved. Opened bug 253592 to deal with new patch request.